Characterization of a GTPase-activating protein that stimulates GTP hydrolysis by both ADP-ribosylation factor (ARF) and ARF-like proteins. Comparison to the ARD1 gap domain.

ADP-ribosylation factors (ARFs) are approximately20-kDa guanine nucleotide-binding proteins that participate in vesicular transport in the Golgi and other intracellular compartments and stimulate cholera toxin ADP-ribosyltransferase activity. Both GTP binding and hydrolysis are necessary for its physiological functions, although purified mammalian ARF lacks detectable GTPase activity. An ARF GTPase-activating protein (GAP) was purified >15,000-fold from rat spleen cytosol using (NH4)2SO4 precipitation and chromatography on Ultrogel AcA 34, DEAE-Sephacel, heparin-Sepharose, hydroxylapatite, and Ultrogel AcA 44. In fractions ( approximately100-kDa proteins) from Ultrogel AcA 44, a major protein band of approximately50 kDa on SDS-polyacrylamide gel electrophoresis correlated with GAP activity, consistent with it being a homodimer, thus differing from an ARF GAP purified from rat liver (Makler, V., Cukierman, E., Rotman, M., Admon, A., and Cassel, D. (1995) J. Biol. Chem. 270, 5232-5237). Purified spleen GAP accelerated hydrolysis of GTP bound to recombinant ARF1, ARF3, ARF5, and ARF6; no effect of NH2-terminal myristoylation was observed. ARF GAP also activated GTP hydrolysis by ARL1, which is 56% identical in amino acid sequence to ARF1, but lacks ARF activity. ARD1 is a 64-kDa guanine nucleotide-binding protein that contains an 18-kDa ARF domain at its carboxyl terminus; the ARF domain lacks the amino-terminal alpha-helix found in native ARF and hence is similar to the amino-terminal truncated mutant Delta13ARF1. Both the ARF domain of ARD1 and Delta13ARF1 were poor substrates for ARF GAP. The non-ARF1 domain of ARD1 enhanced the GTPase activity of the ARF domain, but not that of the ARF proteins and Delta13ARF1, i.e. it lacks the relatively broad substrate specificity exhibited by ARF GAP.

purified on the basis of their ability to increase the ADPribosyltransferase activity of cholera toxin (1)(2)(3). In cells, they are apparently involved in the regulation of exocytic and endocytic vesicular transport pathways (3,4), as well as in the activation of phospholipase D (5,6). ARFs are highly conserved in eukaryotic cells from Giardia to humans; at least six ARFs have been identified in mammalian tissues (7)(8)(9)(10)(11)(12). ARF also occurs as the 18-kDa carboxyl-terminal domain of a 64-kDa protein termed ARD1 (13). The amino-and carboxyl-terminal domains, of ARD1 synthesized in Escherichia coli, interact functionally and the NH 2 -terminal protein stimulates GTPase activity of the ARF-like domain, i.e. serves as its GAP (14). A family of proteins, termed ARLs, for ARF-like proteins, are very similar to ARFs in structure, but lack the ability to activate cholera toxin (15)(16)(17)(18).
ARF proteins are active when GTP is bound. The exchange of bound GDP for GTP, which is necessary to form ARF GTP, is accelerated by a guanine nucleotide-exchange protein or GEP. Membrane-associated GEP activity has been described (19,20), and a GEP from bovine brain cytosol has been partially purified (21). Inactivation results from the hydrolysis of bound GTP, which requires interaction with a GAP or GTPase-activating protein, as ARF itself has no detectable GTPase activity (22)(23)(24). Randazzo and Kahn (23) studied an ARF GAP activity in bovine brain membrane extracts that was stimulated by acidic phospholipids. More recently, the purification (24) and cloning (25) of an ARF GAP from rat liver cytosol has been reported. We describe here the purification and characterization of an ARF GAP from rat spleen cytosol that appears to have a rather broad substrate specificity as it activated GTP hydrolysis by mammalian ARFs from all three classes, with and without myristoylation, as well as by ARF-like ARL proteins.
One-third of the dialyzed solution was applied to a column (5 ϫ 80 cm, 1.57 liters) of Ultrogel AcA 34. During elution with 3 liters of TENDSP buffer containing 0.1 M NaCl (flow rate, 100 ml/h), 15-ml fractions were collected. A single peak of GAP activity was observed; those fractions were combined and concentrated to 30 ml in an Amicon stirred cell (YM-10 membrane).
The concentrated pool from three columns (1500 mg of protein) was dialyzed overnight against TENDSP buffer with 0.5 M sucrose and applied to a column (4 ϫ 28 cm, 352 ml) of DEAE-Sephacel, previously equilibrated with and then washed with 350 ml of TENDSP buffer containing 2 mM MgCl 2 . The column was eluted with 350 ml of TENDSP buffer containing 2 mM MgCl 2 and 0.1 M NaCl, followed by a linear gradient of 100 -240 mM NaCl (850 ml of each) in the same buffer. Fractions containing GAP activity (eluted between 130 and 150 mM NaCl) were pooled, concentrated to ϳ25 ml, and dialyzed as described above.
After dialysis, proteins (220 mg) were applied to a column (30 ml) of heparin-Sepharose, followed by 60 ml of TENDSP buffer with 2 mM MgCl 2 and 20 ml each of the same buffer containing 50 and 100 mM NaCl. On elution with a linear gradient of 100 -350 mM NaCl (110 ml of each) in the same buffer, GAP activity was recovered in a single peak between 200 and 220 mM NaCl.
Pooled peak fractions (17 mg of protein) were applied to a column (1 ϫ 7.6 cm, 6 ml) of hydroxylapatite (HAP), previously equilibrated with TENDSP containing 2 mM MgCl 2 . The column was washed with 6 ml of the same buffer and eluted with a linear gradient of 0 -70 mM potassium phosphate in the same buffer (24 ml of each) followed by a linear gradient of 70 -250 mM potassium phosphate in the same buffer (18 ml of each). GAP activity was collected between 40 and 80 mM phosphate.
Active fractions from HAP were pooled, concentrated to 1.1 ml (Amicon YM 10), and applied to a column (0.9 ϫ 54 cm, 34 ml) of Ultrogel AcA 44, which was equilibrated and eluted with TENDSP buffer containing 2 mM MgCl 2 and 0.1 M NaCl (flow rate, 0.5 ml/3 min). Thirteen active fractions were pooled (total 6.5 ml). ARF GAP was stable at Ϫ20°C for at least 3 months, but activity decreased with repeated freezing and thawing.
Assay of ARF GAP-ARF GAP activity was assayed by its effect on ARF activation of CTA ADP-ribosyltransferase activity or on the hydrolysis of ARF-bound GTP to GDP, quantified after thin layer chromatography (TLC). To measure ARF activation of CTA (assay I), ARF GTP was prepared by incubating 0.2 g of mixed ARFs, 30 g of bovine serum albumin, 10 g of phosphatidylserine, and 5 M GTP in TENDS buffer (total volume 40 l) at 30°C for 2 h before transfer to an ice bath. To the ARF GTP mixture, the ARF GAP preparation (up to 20 l) was added and the total volume adjusted to 100 l with TENDS buffer. After incubation at 30°C for 10 min, mixtures were placed in an ice bath during addition of components of the CTA assay, i.e. 100 l of a solution containing 60 g of ovalbumin, 60 M Cibacron blue, 40 g of phosphatidylserine, and 2 g of CTA plus 100 l of a solution of 150 mM potassium phosphate, pH 7.5, 7.5 mM MgCl 2 , 1.5 mM ATP, 60 mM dithiothreitol, 50 mM agmatine, and 0.6 mM [adenine-14 C]NAD (10 5 cpm)(final volume 300 l). After incubation at 30°C for 60 min, [ 14 C]ADP-ribosylagmatine was isolated for radioassay (2). ARF activity was expressed as the increase in CTA activity (product formed) due to its addition. ARF activity was, of course, decreased by GAP-catalyzed enhancement of GTP hydrolysis.
The TLC-based assay II for ARF GAP was a modification of that described by Randazzo and Kahn (23). ARF (0.6 -1.2 M) was incubated with 0.5 M [␣-32 P]GTP in 20 mM Tris-Cl, pH 8.0, 1 mM MgCl 2 , 0.1% Triton X-100, 2 mM dithiothreitol, and 30 g of bovine serum albumin (total volume: 60 l) at 30°C for 30 min (recombinant ARFs) or 2 h (native ARFs). Samples (10 l) were then diluted with 40 l of the same buffer containing 130 M PIP 2 with or without ARF GAP (final volume: 60 l) and incubated for 10 min at 30°C before dilution with 2 ml of ice-cold 20 mM Tris-Cl, pH 8.0, 100 mM NaCl, 5 mM MgCl 2 , and 2 mM dithiothreitol. Protein-bound nucleotides were collected on nitrocellulose (23) and eluted in 250 l of 2 M formic acid. Samples of eluates were applied to polyethyleneimine-cellulose plates (23) and the remainder was used for radioassay to quantify total nucleotide. Plates were developed with 1 M LiCl/1 M formic acid to separate [␣-32 P]GTP and [␣-32 P]GDP, which were quantified using a PhosphorImager (Molecular Dynamics). Because PIP 2 (used in the GAP assay) causes dissociation of GDP from ARF (28), activity was calculated as ⌬GTP/GTP o , where ⌬GTP is the difference between the amounts of [␣-32 P]GTP bound after incubation without and with GAP, i.e. activity is the fractional decrease during the assay in bound GTP due to the presence of GAP. A unit of GAP activity is defined as the amount causing hydrolysis of 50% of GTP bound to 0.24 g of mixed ARFs in 10 min at 30°C.

RESULTS AND DISCUSSION
For initial experiments, GAP assay I, based on ARF activation of CTA-catalyzed ADP-ribosylagmatine synthesis, was used. This assay relies on the specificity of the ARF GAP interaction and, therefore, does not have a "background" due to non-ARF GTP-binding proteins, a particular problem in assay II with crude GAP preparations. The more widely used assay II allows direct visualization of the substrate and product nucleotides as well as their quantification using a PhosphorImager. This assay was used for the later stages of purification and for characterization of GAP activity.
Results of a representative GAP purification as described under "Experimental Procedures" are summarized in Table I. Crude cytosol and (NH 4 ) 2 SO 4 precipitate data are not included, because assay II was not used for those fractions. Purification was ϳ900-fold from the Ultrogel AcA 34 chromatography product with only 0.8% recovery. Overall purification from cytosol was estimated to be Ͼ15,000-fold.
The purified GAP activity eluted in a symmetrical peak from Ultrogel AcA 44 with an apparent size of ϳ100 kDa (Fig. 1). Analysis of fractions by SDS-polyacrylamide gel electrophoresis in 8% gels followed by silver staining revealed a polypeptide of ϳ50 kDa that co-eluted with activity ( Fig. 1, lower panel). These observations are consistent with the notion that the GAP from rat spleen cytosol is a dimer. This differs from the conclusion of Makler et al. (24) that the GAP, which they purified from rat liver cytosol, is a homotetramer. Although the two GAPs could represent products of the same gene in forms that differ because of purification procedures or tissue sources, they could as well be as quite different proteins.
Both native ARFs 1 and 3 were good substrates for ARF GAP purified through hydroxylapatite (Fig. 2). GAP activity was likewise similar with mixed native ARF and rARF1 (Fig. 3A), whereas rARF5 and rARF6 seemed to be somewhat better substrates than the mixed ARF (Fig. 3, B and C). The myristoylated and non-myristoylated recombinant ARF proteins behaved similarly as GAP substrates (Fig. 3). The interpretation of these negative findings is, however, limited by the lack of quantitative data on the extent of myristoylation and the amount of native active proteins in the recombinant preparations.
Randazzo and Kahn (23) reported that ARF GAP from bovine brain membranes was markedly activated by phosphoinositides, especially PIP 2 . Similar effects of PIP 2 were observed with the partially purified GAP from spleen and with native ARF 1 or ARF 3 (Fig. 4), although they were maximal at lower TABLE I Purification of GAP from rat spleen cytosol These data were obtained using assay II, which did not provide accurate activities for GAP in homogenate, supernatant, or (NH 4 ) 2 SO 4 precipitate. These initial steps provided a Ͼ15-fold purification. Assay I was used to monitor activity in those preparations. The procedure has been replicated three times. concentrations of PIP 2 and of lesser magnitude than those described for the brain GAP. Effects of PIP 2 on the ARF GAP purified from liver cytosol (24) were also smaller than those reported by Randazzo and Kahn (23). It was notable that the magnitude of the effects of PIP 2 (and of mixed phosphatidylinositols) in the studies of Makler et al. (24) appeared to vary with the amount of GAP in the assay and its purity. Although it is difficult with the information available, to evaluate conclusively the physiological significance of PIP 2 effects on GAP activity, it appears that PIP 2 may well have a role in more than one aspect of ARF function. The notable enhancement of ARF activation of phospholipase D by PIP 2 has been well documented (31, 32). As reported by Terui et al. (33), PIP 2 appeared to accelerate release of GDP from ARF. Dissociation of GDP from ARF3 was faster than from ARF1, as seen, for example, in the inset in Fig.  2. Although the amount of bound [ 35 S]GTP␥S remained constant during incubation with GAP, the total amount of bound [␣-32 P]GTP plus [␣-32 P]GDP declined as GTP hydrolysis proceeded (Fig. 5), presumably due to dissociation of bound [␣-32 P]GDP. As effects of differing assay conditions on bound GTP were not observed, GAP activity was expressed as the fractional decrease in bound GTP.
ARF GAP also enhanced the GTPase activity of the ARF-like ARL1 and mutant ARF1 lacking 13 amino acids at the NH 2 terminus (Fig. 6), albeit apparently less effectively than it enhanced the GTPase activity of the native or recombinant ARFs. Myristoylation of the NH 2 -terminal deletion mutant ⌬13ARF1 (Fig. 6) had no apparent effect on its activity, although the interpretation of the observation is necessarily limited, as it is for the data with the recombinant intact ARF proteins with and without myristoylation. As reported (15), the intrinsic GTPase activity of the recombinant ARL1 (non-myristoylated) was higher than that of the ARFs (data not shown). Enhancement by ARF GAP was, however, easily demonstrated (Fig. 6). It seems likely that ARL activity, like that of ARF, is regulated by a GAP protein(s), whether or not by an ARLspecific GAP remains to be demonstrated. It is likewise unclear at present how many different ARF GAPs may exist. ARF GAP is alternatively spliced, and this processing may give rise to proteins with different substrate specificities and responses to phospholipids (25). ARD1 was initially identified by cDNA cloning as a 64-kDa protein with an ϳ18-kDa ARF sequence at the COOH terminus (13). The NH 2 -terminal part of ARD1 (p5) has been shown recently to function as a GAP for the COOH-terminal ARF domain (p3) (14). The ability of p5 to enhance the GTPase activity of ARL1 and several ARF proteins was, therefore, investigated (Table II). Although recombinant p5 markedly increased GTP hydrolysis by p3, as already reported (14), no GAP activity toward native or recombinant ARF proteins or ARL1 was detected. It appears that the NH 2 -terminal portion of ARD1 serves specifically as a GAP for the ARF domain of that protein. This is in marked contrast to the apparently rather broad substrate specificity of the ARF GAP purified from spleen cytosol. Consistent with these data, amino acid sequences of the GAP domain of ARD1 (13) and of peptides from spleen ARF GAP 2 were not notably similar, nor were those of the peptides and the deduced amino acid sequence of a rat liver ARF GAP cloned by Cukierman et al. (25). The meaning of this, at the moment, seemingly negative, information will remain unclear until we are able to demonstrate directly that the peptide sequence is part of a protein that exhibits GAP activity.  P]GTP bound (60 l) was incubated with p5 (12.5 g) for 60 min at room temperature (total volume: 120 l). GTPase activity is expressed as the percentage decrease in bound GTP during incubation for 60 min with GAP (⌬GTP/GTP o ϫ 100), based on PhosphorImager quantification. Data are means of duplicates Ϯ onehalf the range. Each experiment has been repeated at least once.